Epic Emmie

What is Epic Emmie?

Epic Emmie is the AI assistant Epic builds into MyChart for patients. Epic says Emmie answers patients' questions, surfaces information from their charts and takes action for them, and that it can reference chart content such as notes from a recent visit, lab results and past messages. Health systems decide whether and how to offer it.

Emmie belongs to a family of named AI features. In Epic's announcements, Art is the AI for clinicians, which includes AI Charting, and Emmie is the AI for patients. Each health system switches these features on through its own Epic team.

What can patients ask Emmie to do?

Patients can ask Emmie to explain their information and to complete common tasks. Epic's February 2026 update and its Emmie page describe the scope.

• Understand the record: Emmie helps patients navigate MyChart, ask questions and interpret the clinical information available in their account.
• Scheduling: Patients can schedule appointments, and text messaging with conversational AI supports scheduling and rescheduling.
• Billing: Emmie helps patients understand bills, make payments, set up payment plans and generate detailed statements for reimbursement, and can explain why a patient owes what they owe.
• Handoff: When needed, Emmie can escalate a patient to a live chat with a staff member.

Health systems are adopting it one by one. In August 2026, Epic reported that Community Health Network had launched Emmie in MyChart so patients can ask questions about their health records and get answers based on what is in their chart.

Where do Emmie's answers come from?

Emmie's answers come from the patient's own chart and account in Epic. That grounding is the main difference between Emmie and a general chatbot that knows nothing about the patient.

Grounding in the chart also sets the limits. Emmie can only explain what the record holds and what the patient's account can see, so a result that is not yet released to MyChart, or data held in an outside system, is outside its reach. Patients who use proxy access for a child or parent raise the same question, and the health system's MyChart access rules decide it.

For an outside vendor, this matters in two ways. Data your product holds outside Epic is invisible to Emmie unless it is filed into the chart, and anything you file can become part of what patients ask about.

Documentation quality matters more once an assistant explains the chart. Abbreviations, copied text and stale problem lists were always a clinical issue, and they now shape the answers a patient reads in plain language.

Emmie or your own patient assistant?

For questions about the Epic record, scheduling and Epic billing, the health system's own Emmie is usually the first option to evaluate, because it already sits in MyChart with the patient's context. A separate assistant earns its place only where it does something Emmie does not.

• Record questions in MyChart: Evaluate Emmie first. It already has the chart context and the patient's MyChart session.
• Specialty programs outside Epic: A separate app fits when the program's data lives in your system, such as a home monitoring or fertility program with its own records.
• Non-Epic patients: A program that serves patients of several health systems, on different EHRs, needs its own assistant and its own data access.
• Brand and channel: A health system that wants an assistant in its own app or on its website, outside MyChart, needs to check which channels Emmie supports before building.
• Clinical logic: Any assistant that suggests what a patient should do clinically needs clinical review and a regulatory check, whoever builds it.

Build the comparison against the health system's actual Emmie configuration, not Epic's announcements. Features roll out on Epic's release schedule, and each health system decides which ones it turns on.

What changes for patient support teams?

Support teams move from answering every routine question to handling the ones the assistant hands over. That shift needs planning before launch, not after the first complaints.

• Handoff staffing: Live chat escalations need named staff and hours, and patients need to know what happens outside those hours.
• Billing knowledge: A patient who reaches a person after an AI explanation of their bill expects the agent to see the same account detail.
• Scheduling rules: Self-scheduling through any channel follows the visit types and slots the health system opens, so access teams review those rules first.
• Feedback loop: Staff flag wrong or unhelpful answers to the digital team, who adjust configuration and content.
• Reporting: Track handoffs, unresolved conversations and complaints from launch, so leaders can see where the assistant helps and where it does not.

Outside vendors feel this too. A contact center platform or patient app that sits next to MyChart should share the same escalation paths, so a patient does not get one answer from the assistant and a different one from the call center.

How does a third-party patient app work alongside Emmie?

A third-party patient app connects to the same Epic record through Epic on FHIR, with the patient signing in using their MyChart credentials. It works beside Emmie, not inside it.

• Patient sign-in: A SMART on FHIR standalone launch lets the patient authorize your app with MyChart, and the app receives a token scoped to that patient.
• Reading the record: The certified API under 45 CFR 170.315(g)(10) is a read API for patient and population services, built on USCDI through US Core.
• Writing back: Writes need Epic's own write APIs, such as QuestionnaireResponse create for patient-entered questionnaires and Observation create for vital signs, each enabled by the customer.
• Scheduling: Epic on FHIR lists Appointment find and book operations in STU3, which a customer may enable for self-scheduling apps.

Keep the patient's story consistent. If your app files a reading or a questionnaire, Emmie may later explain it to the same patient, so the data should be coded, dated and labeled the way a clinician would expect to read it.

Decide early which system owns each task. If both your app and MyChart can book the same visit type, the health system must agree which one patients are pointed to, or patients will meet two sets of rules for one appointment.

Rules that apply to patient-facing AI

No single rule covers patient-facing AI. The relevant rules depend on who builds and runs the tool, and what it does.

• FDA: The final Clinical Decision Support Software guidance, issued in January 2026, keeps patient-facing tools outside the CDS carve-out, though they are not automatically devices; the general wellness and enforcement discretion policies may apply.
• HIPAA business associates: A vendor that builds or runs a patient tool for a provider and handles PHI on its behalf is a business associate under 45 CFR 164.502(e). An app the patient chooses under the right of access is not the provider's business associate.
• FTC: A health app outside HIPAA falls under the FTC Health Breach Notification Rule, 16 CFR Part 318.
• Accessibility: The HHS Section 504 rule at 45 CFR 84.84 adopts WCAG 2.1 Level AA for the web content and mobile apps of recipients of HHS federal financial assistance, with compliance dates of 11 May 2027 for recipients with 15 or more employees and 10 May 2028 for smaller ones.

A health system that builds its own assistant on a commercial model also needs a BAA from the model provider. OpenAI, for example, signs BAAs for eligible API services with zero data retention, while consumer ChatGPT is not covered.

What to test before patients see AI answers

Test any patient-facing assistant, Epic's or your own, against the questions patients really ask, with clinical and patient-experience reviewers in the loop.

• Grounding: Every answer about the patient traces to a record entry the patient can see.
• Released results only: The assistant never explains a result that has not been released to the patient.
• Proxy accounts: Answers for a parent or caregiver respect the proxy's access level.
• Escalation: Symptoms that need urgent care send the patient to a person or emergency guidance, not to more chat.
• Reading level: Answers are readable for the health system's patient population, and tested in each language offered.
• Accessibility: The chat works with screen readers and keyboard navigation to WCAG 2.1 AA.
• Audit: Each conversation is logged so staff can review what was said when a patient complains.

Keep the test set and run it again after every Epic upgrade and every configuration change. An assistant that answered well at launch can drift when new features, content or data sources are switched on.

How Agnotic can help

When a health system turns on Emmie, its outside patient apps need a clear job. Agnotic Technologies helps define that job and builds the software: patient apps that sign in with MyChart through SMART on FHIR, write-backs that file questionnaires and readings the way clinicians expect, and assistants for programs whose data sits outside Epic.

Agnotic designs patient-facing tools to WCAG 2.1 AA, builds to HIPAA Security Rule safeguards, and works with the customer's Epic team so outside apps and MyChart tell the patient the same story.

Sources

• Epic: Emmie, AI for patients
• Epic (Feb 2026): Epic AI Charting rolls out alongside expanding built-in AI capabilities
• Epic (Aug 2026): Community Health Network adopts Emmie in MyChart
• open.epic: FHIR interfaces, including QuestionnaireResponse, Observation and Appointment
• ASTP/ONC: Standardized API for patient and population services, 170.315(g)(10)
• Federal Register (11 May 2026): Extension of Section 504 web and mobile accessibility compliance dates
• OpenAI: Business Associate Agreement for API services
• 45 CFR 164.502: Uses and disclosures, including business associates